Method and system for object recognition search

Abstract

A method for object recognition using shape and color features of the object to be recognized. An adaptive architecture is used to recognize and adapt the shape and color features for moving objects to enable object recognition.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application is related to and claims the benefit of the following copending and commonly assigned U.S. Provisional Patent Application: U.S. Patent Application No. 61 / 186,304, titled “Shape and Color Features for Object Recognition Search,” filed on Jun. 11, 2009; the entire contents of which is incorporated herein by reference.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT[0002]The invention described herein was made in the performance of work under a NASA contract, and is subject to the provisions of Public Law 96-517 (35 USC 202) in which the Contractor has elected to retain title.BACKGROUND[0003]1. Field[0004]This disclosure relates to object recognition search. More particularly, the present disclosure describes feature extraction techniques that are used to facilitate object recognition and search. The present disclosure also describes architectures and systems that are particularly adapted for the recognition a...

AI Technical Summary

Benefits of technology

The present invention is a real-time adaptive visual system that uses bio-inspired features, adaptive architecture, and autonomous mechanisms. It extracts knowledge of color and shape features from a statistical image collection and stores it in databases. The system then detects shape and color features in multiple images and identifies objects within them by matching the stored features. The system can also extract new features from the identified objects and use them to recognize the object in other images. The technical effect of this invention is an efficient and accurate method for identifying objects in multiple images.

Problems solved by technology

While humans are well-adapted for object recognition, machine or computer-based object recognition is a challenging problem.

For problems with rich data sources such as image recognition and computer vision, the dimension of the input vector is typically larger than the number of input samples, which leads to overfitting of the data if no care is taken to achieve useful generalization.

Furthermore, computing power rapidly becomes insufficient to process the data within a reasonable time.

Unfortunately, PCA requires that the basis vectors be orthogonal, which is typically an artificial assumption.

The PCA procedure is complicated and computationally intensive (O(N3), where N is the dimension of the vector input), thereby making it difficult to use for rich data sources.

For sequential PCA, the gradient descent (GED) technique may be a more attractive approach for hardware implementation as a straight forward technique compared to others, e.g., steepest decent, conjugate gradient, or Newton's second order method, but it exposes some difficulties in learning convergence when other principal component vectors are corresponding to smaller eigen values.

In addition, this technique still requires some complicated hardware.

While techniques are known in the art for object detection and recognition based on either shape feature extraction and detection or color feature extraction and detection, the usefulness of these techniques for use in a heterogeneous environment may be somewhat limited.

Images